Explanation Regularisation through the Lens of Attributions
Not on the map. Why?
Abstract (arXiv)
Explanation regularisation (ER) has been introduced as a way to guide text classifiers to form their predictions relying on input tokens that humans consider plausible. This is achieved by introducing an auxiliary explanation loss that measures how well the output of an input attribution technique for the model agrees with human-annotated rationales. The guidance appears to benefit performance in out-of-domain (OOD) settings, presumably due to an increased reliance on "plausible" tokens. However, previous work has under-explored the impact of guidance on that reliance, particularly when reliance is measured using attribution techniques different from those used to guide the model. In this work, we seek to close this gap, and also explore the relationship between reliance on plausible features and OOD performance. We find that the connection between ER and the ability of a classifier to rely on plausible features has been overstated and that a stronger reliance on plausible tokens does not seem to be the cause for OOD improvements.
Key facts. Cited by 1 paper in the map (data through 14 Oct 2026). Builds on 47 of its 62 references that are in the map. First posted on arXiv 23 Jul 2024.
Builds on
47 of 62 references.
References · 47
- 2017Attention Is All You Needcited by 60,361
- 2019PyTorch: An Imperative Style, High-Performance Deep Learning Librarycited by 20,305
- 2017Axiomatic Attribution for Deep Networkscited by 3,384
- 2021Convex optimizationcited by 2,515
- 2015Character-level Convolutional Networks for Text Classificationcited by 2,330
- 2020Big Bird: Transformers for Longer Sequencescited by 1,684
- 2017Learning Important Features Through Propagating Activation Differencescited by 1,458
- 2019What Does BERT Look At? An Analysis of BERT's Attentioncited by 1,003
- 2019Attention is not Explanationcited by 929
- 2020Quantifying Attention Flow in Transformerscited by 775
- 2019Attention is not not Explanationcited by 636
- 2020Transformer Interpretability Beyond Attention Visualizationcited by 627
35 more
- 2020Captum: A unified and generic model interpretability library for PyTorchcited by 547
- 2019ERASER: A Benchmark to Evaluate Rationalized NLP Modelscited by 530
- 2018e-SNLI: Natural Language Inference with Natural Language Explanationscited by 468
- 2020HateXplain: A Benchmark Dataset for Explainable Hate Speech Detectioncited by 407
- 2019Is Attention Interpretable?cited by 389
- 2019Explain Yourself! Leveraging Language Models for Commonsense Reasoningcited by 367
- 2017Right for the Right Reasons: Training Differentiable Models by Constraining their Explanationscited by 294
- 2024Bridging Language and Items for Retrieval and Recommendation: Benchmarking LLMs as Semantic Encoderscited by 224
- 2019Interpretations are useful: penalizing explanations to align neural networks with prior knowledgecited by 145
- 2019Learning to Deceive with Attention-Based Explanationscited by 111
- 2020Evaluating Explanations: How much do explanations from the teacher aid students?cited by 94
- 2020Aligning Faithful Interpretations with their Social Attributioncited by 79
- 2021When Can Models Learn From Explanations? A Formal Framework for Understanding the Roles of Explanation Datacited by 75
- 2019Incorporating Priors with Feature Attribution on Text Classificationcited by 68
- 2021Improving Deep Learning Interpretability by Saliency Guided Trainingcited by 68
- 2022Measuring the Mixing of Contextual Information in the Transformercited by 65
- 2016Investigating the influence of noise and distractors on the interpretation of neural networkscited by 54
- 2021Incorporating Residual and Normalization Layers into Analysis of Masked Language Modelscited by 51
- 2022GlobEnc: Quantifying Global Token Attribution by Incorporating the Whole Encoder Layer in Transformerscited by 44
- 2022Optimizing Relevance Maps of Vision Transformers Improves Robustnesscited by 41
- 2020Evaluating and Characterizing Human Rationalescited by 36
- 2019Regularizing Black-box Models for Improved Interpretabilitycited by 35
- 2021UNIREX: A Unified Learning Framework for Language Model Rationale Extractioncited by 32
- 2023DecompX: Explaining Transformers Decisions by Propagating Token Decompositioncited by 31
- 2021The Eval4NLP Shared Task on Explainable Quality Estimation: Overview and Resultscited by 27
- 2021What to Learn, and How: Toward Effective Learning from Rationalescited by 25
- 2019Saliency Learning: Teaching the Model Where to Pay Attentioncited by 23
- 2021Translation Error Detection as Rationale Extractioncited by 22
- 2022Learning to Scaffold: Optimizing Model Explanations for Teachingcited by 20
- 2023Studying How to Efficiently and Effectively Guide Models with Explanationscited by 16
- 2022A survey on improving NLP models with human explanationscited by 15
- 2022VisFIS: Visual Feature Importance Supervision with Right-for-the-Right-Reason Objectivescited by 12
- 2022ER-Test: Evaluating Explanation Regularization Methods for Language Modelscited by 6
- 2024Exploring the Trade-off Between Model Performance and Explanation Plausibility of Text Classifiers Using Human Rationalescited by 5
- 2023REFER: An End-to-end Rationale Extraction Framework for Explanation Regularizationcited by 4
Built on by
Citing papers per year: 2025: 1.
1 paper.